Contract Contingencies and Uncertainty: Evidence from Product Market Contracts

Kai Wai Hui
Jun Oh
Guoman She
P. Eric Yeung

2025-04-28

An example of contract contingency

  • Product delivery shall be made within 30 days of the cash deposit.
  • Product delivery shall be made within 30 days of the cash deposit. In the event of shipping disruptions, the delivery may be further delayed by no more than 90 days.
  • Contract contingency refers contigencies written in contracts.

Research question

  • How does uncertainty affect contract contingencies?

Table of contents

Motivation

Existing literature sutdies the contract completeness

  • More complete contracts have lower ex-post opportunistic behavior but bears higher ex-ante writing costs (e.g., Crocker and Reynolds 1993).

Existing literature usually relies on a small sample from one industry

  • Crocker and Reynolds (1993) examine 44 contracts in the U.S. defense industry;
  • Saussier (2000) examines 29 contracts in the French coal industry;
  • Anderson and Dekker (2005) examine 858 contracts in the information technologies (IT) industry;
  • Aubert et al. (2017) use 173 survey observations regarding outsourcing contracts in the IT industry;
  • Krishnan and Mani (2020) analyze 455 strategic outsourcing contracts.

Contract contingency is a better perspective to study contract completeness

  • Existing literature often assumes that more actions included in a contract (or a longer contract) indicate more complete contracts.
  • As opposed to contract completeness, this study focuses on well-defined contract contingencies.

“clause comprehensiveness” → “risk response capability”

Theoretical Framework

Writing cost perspective

  • In the traditional contracting theory framework, writing contingencies is costless (Arrow 1964).
  • With bounded rationality and opportunism:
    • thinking about possible outcomes;
    • seeing through their implications;
    • designing (and bargaining) related actions;
    • and writing them in a contract formally are all costly activities (Dye 1985).
  • Dye (1985) argues that writing cost is propotional to the number of contingencies.
  • As uncertainty grows, too cost-ineffective to:
    • contemplate all possible outcomes, each with a small likelihood,
    • bargain and negotiate corresponding actions.

graph TD
    classDef dashed fill:none,stroke-dasharray: 5 5;
    subgraph "High Uncertainty"
        direction TB
        T0h2[T=0] -->|20%| A1h2[Action 1]
        T0h2 -->|5%| A2h2[Action 2]
        T0h2 -->|5%| A3h2[Action 3]
        T0h2 -.-> Ax1[Action X1]
        T0h2 -.-> Ax2[Action X2]
        Ax1:::dashed
        Ax2:::dashed
    end

    subgraph "Low Uncertainty"
        direction TB
        T0h1[T=0] -->|60%| A1h1[Action 1]
        T0h1 -->|20%| A2h1[Action 2]
        T0h1 -->|20%| A3h1[Action 3]
    end

  • High uncertainty → fewer contingencies
  • Since with fewer contigencies, parties have to amend actions upon observing the outcome realization.

Dynamic contracting models

  • Conditional on the written contingency, the cost of amending actions ex-post is much lower than re-negotiating an entirely new contract (Battigalli and Maggi, 2008).
  • Alternatively, actions drafted roughly in the first place often serve as good reference points to facilitate future amendments (Hart and Moore 2008).
    • Example: a “normal” shipping disruption-related contingency helps serve as a good reference point for determining the period of further delay under more “severe” shipping disruptions.
  • Roughly drafted contingencies lower ex-ante writing costs while facilitating less costly future amendments (Tirole, 2009).
  • High uncertainty: limited number of roughly drafted clauses + ex-post amendments(Tirole, 2009).
  • High uncertainty → more contingencies (roughly drafted)

Price protection view

  • Williamson (1985): If the parties could ex-ante negotiate payoffs to provide protection, a contract with few contingencies would still be deployed under high uncertainty.
  • Grossman and Hart (1986): The party with higher exposure retains asset ownership and bears the residual risks.
  • Maskin and Tirole (1999): Uncertainty does not matter if the parties can probabilistically forecast future payoffs.
  • No significant relationship between uncertainty and contingency

Data and Descriptive Statistics

Sample and measurement of contigencies

  • Sample period: 1996 - 2021
Sample selection procedure
All product market contracts 4553
Less: Contracts by filers unidentified within compustat (293)
Less: Filers without SIC industry code (73)
Less: Contracts with insufficient text data to measure
contingencies (75)
Less: Contracts without firm-level control variables (106)
Main sample with ContingencyR as dependent variable 4006
  • Identify relevant subsections: develop an algorithm to split a contract into subsections.
  • Group subsections into “topics” (i.e., delivery, warranty): using LDA model.
  • Identify actions and contingencies on topic level: a topic is designated a “contingency topic” if it contains at least one contingency clause.
  • Aggregate actions and contingencies to the contract level:
  • ContingencyR = contingency topics / action topics

Measurement of general uncertainty

  • Industry earnings volatility
    • industry-level average earnings volatility
  • Industry cash flow volatility
    • industry-level average cash flow volatility
  • Industry return volatility
    • industry-level average stock return volatility

Descriptive statistics

Contract-level variables
Variable Mean Std P25 P50 P75
Words in contract 1558.554 1299.351 720.000 1260.500 2058.000
Subsections per contract 73.196 57.005 36.000 61.000 96.000
Topic_Action 26.464 11.588 18.000 27.000 35.000
Topic_Contingency 13.305 7.466 8.000 13.000 19.000
ContingencyR 0.479 0.133 0.400 0.500 0.568
InnoTopic_Action 5.876 4.977 2.000 5.000 8.000
InnoTopic_Contingency 2.561 2.871 1.000 2.000 4.000
InnoContingencyR 0.391 0.290 0.167 0.400 0.571
Uncertainty variables
Variable Mean Std P25 P50 P75
Industry earnings volatility 0.162 0.082 0.095 0.162 0.232
Industry cash flow volatility 0.128 0.065 0.075 0.117 0.181
Industry return volatility 0.157 0.044 0.126 0.166 0.190
Industry R&D intensity 0.134 0.108 0.019 0.123 0.240
Industry patent originality 0.721 0.496 0.322 0.710 1.000
Industry trade secrecy 0.671 0.262 0.474 0.764 0.894

Research Design and Empirical Findings

Research design

\[ Contingency_c = \beta \ IndustryUncertainty_{j,t} + X_{i,j,t - 1} + \gamma_t + \nu_j + \epsilon_c \]
where c, i, j, and t index contract, firm, industry, and year.

  • X includes
    • industry ROA, firm ROA, firm gross profit margin, firm sales, firm assets, number of firm segments, firm age, industry average of market-to-book ratio, firm leverage ratios, firm trade credit, same state, private.
  • Control year and industry FEs.

Main results

General uncertainty and contract contingency
Dep. Var. = ContingencyR (1) (2) (3)
Industry earnings volatility 0.1147
[2.81]***
Industry cash flow volatility 0.1240
[2.13]**
Industry return volatility 0.1880
[2.08]**
Controls Yes Yes Yes
Observations 4006 4006 4006
Adjusted R - squared 0.0767 0.0761 0.0765
Sector & year FE Yes Yes Yes
Innovation uncertainty and contract contingency
Dep. Var. = ContingencyR (1) (2) (3)
Industry R&D intensity 0.1840
[5.28]***
Industry patent originality 0.0193
[2.07]**
Industry trade secrecy 0.0346
[1.81]*
Controls Yes Yes Yes
Observations 4006 3702 4006
Adjusted R - squared 0.0797 0.0793 0.0761
Sector & year FE Yes Yes Yes
Innovation uncertainty and innovation-related contingency
Dep. Var. = InnoContingencyR (1) (2) (3)
Industry R&D intensity 0.5135
[4.09]***
Industry patent originality 0.0510
[3.03]***
Industry trade secrecy 0.2208
[3.19]***
Controls Yes Yes Yes
Observations 3167 2944 3167
Adjusted R - squared 0.0396 0.0384 0.0409
Sector & year FE Yes Yes Yes

Identification strategy

  • General uncertainty shocks: Financial crisis and COVID-19 pandemic \[ \begin{aligned} \text{ContingencyR}_{c} &= \beta_{1}\text{PostFinCrisis}_{t}+\beta_{2}\text{HighLev}_{i}\\ &+ \beta_{3}\text{PostFinCrisis}_{t} \times \text{HighLev}_{i}+X_{i,j,t - 1}\\ &+ \sum \beta_{n}\text{PostFinCrisis}_{t} \times \text{Performance}_{i,j}+v_{j}+\epsilon_{c} \end{aligned} \] \[ \begin{aligned} \text{ContingencyR}_{c} &= \beta_{1}\text{PostCovid}_{t}+\beta_{2}\text{HighCovidExp}_{i}\\ &+ \beta_{3}\text{PostCovid}_{t} \times \text{HighCovidExp}_{i}+X_{i,j,t - 1}\\ &+ \sum \beta_{n}\text{PostCovid}_{t} \times \text{Performance}_{i,j}+\gamma_{t}+v_{j}+\epsilon_{c} \end{aligned} \]
Increase in general uncertainty (Part 1)
Dep. Var. = ContingencyR (1) (2)
Post_FinCrisis 0.0386
[3.25]***
0.0153
[0.93]
HighLev −0.0075
[-0.44]
Post_FinCrisis × HighLev 0.0373
[1.75]*
Controls Yes Yes
Post_FinCrisis x performance Yes Yes
Observations 283 283
Adjusted R-squared 0.2844 0.2817
Sector FE Yes Yes
Increase in general uncertainty (Part 2)
Dep. Var. = ContingencyR (1) (2)
Post_Covid 0.1034
[4.29]***
−0.0113
[-0.36]
HighCovidExp −0.0113
[-0.42]
Post_Covid × HighCovidExp 0.1260
[2.78]**
Controls Yes Yes
Post_Covid x performance Yes Yes
Observations 134 134
Adjusted R-squared 0.0176 0.0730
Sector FE Yes Yes
  • Innovation uncertainty shock: Industry technology diffusion
    • 29 disruptive technologies diffusion following Bloom et al. (2021)
    • Industry Technology Diffusion is the decile-ranked exposure of the firm’s industry to the 29 disruptive technologies. \[ \begin{aligned} \text{ContingencyR}_{c} =& \beta\text{Industry Technology Diffusion}_{j, t}\\ &+X_{i,j,t - 1}+\gamma_{t}+v_{j}+\epsilon_{c} \end{aligned} \]
Increase in innovation uncertainty
(1)
ContingencyR
(2)
InnoContingencyR
Industry technology diffusion 0.0439
[3.44]***
0.1012
[3.86]***
Controls Yes Yes
Observations 1501 1254
Adjusted R - squared 0.0731 0.0667
Sector & year FE Yes Yes

Further evidence of dynamic contracting model

  • Expected re-negotiation costs and the contingency-uncertainty relation
    • Dynamic contracting models: more costly to re-negotiate a new contract than amending actions ex-post conditional on the written contingency.
      • Product market rivals: more costly to re-negotiate;
      • More falmiliar contracting parties: less costly to re-negotiate;
      • More important contracting parties: less costly to re-negotiate.
General uncertainty and contingency
Dep. Var. = ContingencyR (1)
Rival
(2)
Relation
(3)
Major Customer
High industry earnings volatility 0.0079
[1.07]
0.0163
[2.07]**
0.0205
[3.48]***
Re-negotiation cost -0.0205
[-1.03]
0.0217
[1.67]*
0.0037
[0.26]
High industry earnings volatility × Re-negotiation cost 0.0517
[2.60]**
-0.0747
[-3.55]***
-0.0320
[-1.96]*
Controls Yes Yes Yes
Observations 4006 4006 1675
Adjusted R - squared 0.0787 0.0778 0.1049
Sector & year FE Yes Yes Yes
  • Simlar findings in contingency-innovation uncertainty relations
  • Roughly drafted contingencies
    • More roughly drafted contingencies under high uncertainty according to the dynamic contracting models.
Dep. Var. = Log(Length/#Contingencies) (1)
All Clauses
(2)
All Clauses
(3)
Innovation Clauses
Industry earnings volatility -0.0352
[-2.51]**
Industry R&D intensity -0.0535
[-2.22]**
-0.1100
[-2.09]**
Controls Yes Yes Yes
Observations 72286 72286 10158
Adjusted R-squared 0.1162 0.1162 0.1717
Topic, sector & year FE Yes Yes Yes
  • No impact on length of normal action clauses under high uncertainty.
  • Non-linearity
    • Could there be a threshold?
Non-Linearity test (General uncertainty)
Dep. Var. = ContingencyR (1)
Industry earnings volatility (P10_P20) 0.0026
[0.17]
Industry earnings volatility (P20_P30) 0.0198
[2.08]**
Industry earnings volatility (P30_P40) 0.0104
[0.81]
Industry earnings volatility (P40_P50) 0.0283
[2.49]**
Industry earnings volatility (P50_P60) 0.0217
[1.49]
Industry earnings volatility (P60_P70) 0.0241
[2.77]***
Industry earnings volatility (P70_P80) 0.0253
[3.03]***
Industry earnings volatility (P80_P90) 0.0282
[2.77]***
Industry earnings volatility (P90_P100) 0.0273
[2.14]**
Controls Yes
Observations 4006
Adjusted R - squared 0.0757
Sector & year FE Yes
  • Similar findings in innovation uncertainty
  • Robustness tests
    • Firm-level measures of uncertainty
    • Alternative measures of industry earnings volatility
    • Alternative measures of innovation uncertainty
    • Controlling for more counterparty characteristics

Conclusion and Potential Contributions

Conclusion

  • This paper documents robust evidence of a positive relation between contract contingencies and business uncertainty, which is consistent with predictions from the dynamic contracting models.

Potential contributions

  • Test incomplete contract theories using a large sample of material product market contracts
    • Existing work tends to focus on smaller samples of product market contracts from specific industries (e.g., Costello, 2013; Krishnan and Mani, 2020).
  • Offer the first empirical study of contract contingency and provide novel evidence of its determinants
    • This new contracting feature represents a major advancement from prior studies’ focus on the number of actions to proxy for “contract completeness”.
  • Broaden the scope of empirical accounting research in contracting (which focuses heavily on financial contracts and executive compensation contracts)
    • Product market contracts are fundamentally different from financial contracts, as the former cover a broad spectrum of operating and investing activities while the latter focus on financing activities.

Q&A